Tyler Cowen has a charactestically fine piece in yesterday’s NYT, with much good sense in it. Among other things, however, he says this:

The ad hoc aspect of the bailout created a precedent for what has come to be called “regulation by deal” — now the government’s modus operandi. Rather than publicizing definite standards and expectations for bailouts in advance, the Fed and the Treasury confront each particular crisis anew. Decisions are made as to whether a merger is possible, whether a consortium can be organized, what kind of loan guarantees can be offered and what kind of concessions will be extracted in return. So far, every deal — or lack thereof, in the case of Lehman Brothers — has been different.

While there are some advantages to leaving discretion in regulators’ hands, this hasn’t worked out very well. It has become increasingly apparent that the market doesn’t know what to expect and that many financial institutions are sitting on the sidelines, waiting to see what regulators will do next. Regulatory uncertainty is stifling the ability of financial markets to engineer at least a partial recovery.

But earlier on, he says this:

With the Long-Term Capital bailout as a precedent, creditors came to believe that their loans to unsound financial institutions would be made good by the Fed — as long as the collapse of those institutions would threaten the global credit system. Bolstered by this sense of security, bad loans mushroomed.

So clearly a single act, albeit an exercise of administrative discretion, can create expectations and influence behavior. The problem, then, must be that that the government has not acted consistently since then. The inconsistency has created uncertainty that has frozen the credit market. But is that really true? Has the government really acted inconsistently? And, if so (with Lehman the primary exhibit), didn’t most of the government actions occur after the onset of the crisis? Doesn’t Tyler’s argument imply that the government should have bailed out Lehman, as that would have enhanced consistency and hence predictability? Yet I think he believes the contrary.

Steven Davidoff and David Zaring coined the “regulation by deal” phrase in a recent paper that traces the government’s response to the financial crisis. But what exactly does “regulation by deal” mean?

In essence, it refers to cases where the government pays someone to act in a certain way. In the typical bailout, the government gives cash or credit or some other valuable consideration in return for control rights, a right to repayment of some sort if all goes well, and a few other things (lower executive pay, no more private planes, and so forth). The usual form of regulation, by contrast, involves the passing of rules and the punishment of people who violate them.

Regulation by deal is hardly confined to the financial sector. Nearly every criminal case ends in a deal: the government offers a reduced sentence in return for a confession and cooperation. Governments make deals with witnesses when they offer them protection in return for testimony. Governments pay informers, they offer bounties, indeed they buy all sorts of thing—tanks, tractors, land, intellectual property rights, the services of contractors and employees. Most of the time regulation by deal doesn’t bother us; why should it bother us now? In each case, the government pays someone for his consent to alienate something of value.

Tyler complains that the government doesn’t publicize “definite standards and expectations for bailouts in advance.” I’m not sure the government ever publicizes definite standards and expectations in regulation-by-deal situations. Imagine how hard it would be: “We will buy land when…”—when what? When we want it, that’s when. If you want to buy things from people, you usually don’t want to announce the terms in advance—that eliminates your bargaining power and opens up all sorts of avenues for strategic behavior by potential sellers. Prosecutorial discretion has frequently been criticized, and the deals that prosecutors make may not always be consistent, but no one has found a plausible alternative approach.

To be sure, the government sometimes does lay out standards for deals or purchases; indeed, we have some bailout rules in place. The FDIC system provides that if banks become insolvent, the government will pay some of their creditors up to a certain amount. This is plainly insufficient for a real financial crisis—indeed, the point of maintaining regulatory flexibility to address financial crises is that the usual regulatory tools have failed.

The question for Tyler is just what would the bailout rules that he has in mind look like? I suspect that an adequate set of rules could not be invented—and that the lesson of his column is not that regulatory discretion to address crises in the financial sector is wrong, but that a particular exercise of that discretion, the decision to bail out LTC, was wrong. What would the rules be? The institution has to be big, no? But how big? Doesn’t it matter whether the financial health of other firms depend on the firm in question; if so, how many such firms and to what degree? And isn’t the overall health of the economy also a relevant factor—indeed, Tyler points to the different conditions that prevailed in 1998—the budget surplus, the booming economy, the small size of the derivatives market—as a reason why the LTC bailout was a mistake. How should these factors be incorporated into a rule?

The decision to bail out a firm or industry rests on a mixture of economic, psychological, and political imponderables that can’t be reduced to rules. Or such is my view. If this view is wrong, what should the bailout rules be?